AMBER Archive (2006)

Subject: AMBER: problems for running sander.MPI

From: Qizhi Cui (qzcui_at_scripps.edu)
Date: Wed Oct 11 2006 - 15:39:36 CDT


Hi,

We have installed Amber9 on linux cluster (Fedora Core, glibc-2.3.2). We
used compiler, gfortran, with parallel option, -openmpi. The compiling was
successful. But when we run sander.MPI, the program often randomly stops
at some steps with the error message as following:

[enode05:03662] mca_btl_tcp_frag_send: writev failed with errno=104

(enode05 is one of the node names of the cluster.)

We used command:

mpirun -v -hostfile ehost --mca pls_rsh_agent rsh -np 16 /bin/mosrun -h
sander.MPI -O \
....

Normmally, there is no problem for minimization and constant NVT steps.
The problems often occur during constant NPT and production run.

We also tried mpich option.

Any comments or suggestions are highly appreciated!

Thanks,

Qizhi

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu